The AI industry is learning the wrong lessons from Vasily Grossman

Vasily Grossman humanized Soviet Russia — to his peril. That is the central argument of a new essay published in The Baffler on July 2, 2026. The piece, by critic and translator Alexei Fuelling, reads Grossman’s epic novels Life and Fate and Everything Flows as acts of moral witness against a system that demanded submission. Grossman wrote the human scale into a century of mass violence. The Soviet state punished him for it.

The essay is not about AI. It is about totalitarianism, literary courage, and the cost of telling the truth under a regime that treats truth as a weapon. But it arrives at a moment when the AI industry is building systems that operate at a scale Grossman would recognize. The same logic that crushed individual lives into statistical abstractions now powers recommendation algorithms, surveillance infrastructure, and the training data for frontier models.

The Baffler piece is worth reading for its own sake. It is also worth reading as a mirror.

Fuelling traces Grossman’s trajectory from a loyal Soviet journalist who covered the Battle of Stalingrad to a writer who understood that the Soviet system was not a flawed utopia but a machine for destruction. Grossman’s breakthrough, Fuelling argues, was his insistence on the particular. In Life and Fate, a Jewish physicist imprisoned in a Nazi camp and a Soviet scientist working on the atomic bomb are both trapped by the same logic: the state sees them as interchangeable units. Grossman refuses to let them be interchangeable. He gives them names, memories, small gestures of kindness.

The AI industry operates on the opposite principle. Every large language model, every recommendation system, every predictive policing tool depends on treating people as interchangeable data points. The individual is noise. The aggregate is signal. The goal is to find patterns across millions of lives, not to understand one.

This is not a moral equivalence. No one is building a Gulag for AI. But the intellectual habit is the same: the reduction of human experience to a statistical distribution. The Baffler essay does not make this connection. It does not need to. The connection is structural.

Grossman’s great insight, Fuelling writes, was that “the music of destruction” — the phrase comes from Everything Flows — is not a single catastrophic event but a rhythm that becomes normal. The state’s violence is not a series of shocks. It is a pulse. People adapt to it. They learn to hear it as background noise.

The AI industry has its own music of destruction. It is the quiet hum of data centers, the steady churn of content moderation queues, the invisible sorting of people into credit scores, hiring pools, and insurance tiers. The harm is diffuse, cumulative, and normalized. A misclassification here. A denied loan there. A recommendation algorithm that nudges someone toward radicalization over months, not minutes.

The Baffler piece is not a policy paper. It does not propose regulations or technical fixes. It offers something rarer: a vocabulary for describing what is lost when power becomes abstract and unaccountable. Grossman’s word for that loss is “the senselessness of the particular.” When the state decides that your life is a data point, the particular becomes senseless. It has no meaning except as part of a larger pattern.

The AI industry has not solved this problem. It has automated it.

There is a direct line from the Soviet bureaucrat who filled out forms about prisoners to the machine learning engineer who labels training data. Both are doing a job. Both are contributing to a system whose full effects they cannot see. Grossman understood that the bureaucrat is not evil. The bureaucrat is just following procedure. The evil is in the procedure itself.

Fuelling’s essay is a reminder that the most dangerous systems are not the ones that announce themselves as dangerous. They are the ones that present themselves as neutral, technical, inevitable. The Soviet Union called itself a workers’ paradise. The AI industry calls itself a force for democratization. Both claims demand a suspension of disbelief.

The essay closes with Grossman’s death in 1964, his major works unpublished in his lifetime, his manuscripts smuggled out of the Soviet Union on microfilm. He never saw the world he was writing for. He wrote anyway.

The AI industry is building for a world it also cannot see. The models are trained on the past. The deployment is in the present. The consequences will unfold over decades, in ways no one can predict. The question is not whether the models will be accurate. The question is whether the people building them have the moral imagination to understand what they are building.

Grossman had that imagination. He paid for it.

The Baffler essay does not answer the question. It only asks it. That is enough.

The most dangerous systems are not the ones that announce themselves as dangerous. They are the ones that present themselves as neutral, technical, inevitable.

For the AI industry, the uncomfortable truth is that the music of destruction is not a sound you can mute. It is the sound of the system running. The hum of the servers. The click of the labeler. The quiet satisfaction of a model that converges. Grossman heard that sound in the Soviet Union. He wrote against it. He lost.

The AI industry has not lost yet. It has not even decided whether it is fighting.